{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.22413", "kind": "paper", "url": "https://arxiv.org/abs/2605.22413"}, {"id": "gh:wwwt0ri/receiptbench", "kind": "repo", "url": "https://github.com/wwwT0ri/ReceiptBench"}], "authors": ["Yandi Wang", "Libin Zhan", "Ziwei Huang", "Tiancheng Luo", "Yuxuan Jiang", "Wang Dong", "Leilei Gan", "Jun Chen"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.22413", "source_url": "https://arxiv.org/abs/2605.22413"}, "categories": ["General AI", "Reasoning", "Vision & 3D"], "description": {"en": "ReceiptBench evaluates multimodal large language models on visual information extraction from receipts. It includes 10k human-annotated receipts and four hierarchical subtasks: basic perception, format normalization, semantic reasoning, and structure parsing, with defined metrics and scoring."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.22413", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.22413", "title": "ReceiptBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Zhejiang University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2605.22413", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2605.22413", "source_url": "https://openalex.org/W7162111833"}, "key": "claire-radar:2605.22413", "modality": null, "name": "ReceiptBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_receiptbench_9f399a36", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "fdee4be1c0ddc6e2f7590d9a5e9d346b19c53cc01bdffef19a99ae852574b513", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.22413"}, "publisher": null, "released": "2026-05-21", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.22413", "source_url": "https://arxiv.org/abs/2605.22413"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-22413", "source": "claire_radar", "source_benchmark_id": "2605.22413", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Vision & 3D", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/wwwT0ri/ReceiptBench", "githubScope": "benchmark_repo", "githubStars": 4, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2605.22413", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "capabilities": ["Reasoning"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "92f6009abe115058a572d105ce9d8bc1c0ec93cba81e4c0fff9da5794c395390", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "ReceiptBench evaluates multimodal large language models on visual information extraction from receipts. It includes 10k human-annotated receipts and four hierarchical subtasks: basic perception, format normalization, semantic reasoning, and structure parsing, with defined metrics and scoring.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "To bridge this gap, we introduce ReceiptBench, a large-scale, human-annotated benchmark consisting of 10k diverse receipts, organizing information extraction into four hierarchical sub-tasks: (1) Basic Perception for raw text spotting, (2) Format Normalization for strictly following standardization instructions, (3) Semantic Reasoning for inferring implicit attributes from context, and (4) Structure Parsing for handling nested line items."}, "familyId": "bmf_268e28d589d5", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/wwwT0ri/ReceiptBench", "url": "https://github.com/wwwT0ri/ReceiptBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/wwwT0ri/ReceiptBench", "url": "https://huggingface.co/papers/2605.22413"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/wwwT0ri/ReceiptBench", "url": "https://arxiv.org/abs/2605.22413"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/wwwT0ri/ReceiptBench.", "url": "https://arxiv.org/abs/2605.22413"}], "scope": "benchmark_repo", "url": "https://github.com/wwwT0ri/ReceiptBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_receiptbench_9f399a36", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/wwwT0ri/ReceiptBench", "data": null, "hfPaper": "https://huggingface.co/papers/2605.22413", "paper": "https://arxiv.org/abs/2605.22413", "pdf": "https://arxiv.org/pdf/2605.22413", "project": null, "report": "https://arxiv.org/abs/2605.22413"}, "metrics": [], "motivation": "Extracting structured information from visual documents (Visual Information Extraction, VIE) is a cornerstone of business automation.", "name": "ReceiptBench", "oneLine": "ReceiptBench evaluates multimodal large language models on visual information extraction from receipts. It includes 10k human-annotated receipts and four hierarchical subtasks: basic perception, format normalization, semantic reasoning, and structure parsing, with defined metrics and scoring.", "paperTitle": "From Recognition to Reasoning: Benchmarking and Enhancing MLLMs on Real-World Receipt Document Understanding", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.22413", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "T0rl (ReceiptBench team)", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/wwwT0ri/ReceiptBench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-21", "source": {"authors": ["Yandi Wang", "Libin Zhan", "Ziwei Huang", "Tiancheng Luo", "Yuxuan Jiang", "Wang Dong", "Leilei Gan", "Jun Chen"], "categories": ["cs.CV"], "id": "2605.22413", "title": "From Recognition to Reasoning: Benchmarking and Enhancing MLLMs on Real-World Receipt Document Understanding", "type": "arxiv", "url": "https://arxiv.org/abs/2605.22413"}, "sourceUpdatedAt": "2026-05-21T12:37:03Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "Existing VIE benchmarks lack scale, realism, and semantic granularity. ReceiptBench provides a standardized, publicly available evaluation path for comparing models on diverse receipt understanding tasks, supporting practical deployment decisions in document automation."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
